ordinal classification

To address performance degradation caused by label noise in ordinal image classification, this paper proposes ORDAC—the first framework to integrate Label Distribution Learning (LDL) into ordinal noise-robust modeling. ORDAC employs a data-centering adaptive mechanism to dynamically estimate and optimize the mean and standard deviation of each sample’s label distribution, enabling differentiable, fine-grained label correction—rather than discarding suspicious samples—under both symmetric and asymmetric Gaussian noise assumptions. Experiments on the Adience and Diabetic Retinopathy datasets demonstrate significant improvements: on Adience with 40% label noise, the mean absolute error decreases from 0.86 to 0.62, and recall rises from 0.37 to 0.49; moreover, ORDAC effectively mitigates inherent noise present in the original data. By unifying ordinal constraints with distributional label modeling, ORDAC establishes a novel paradigm for noise-robust learning in ordinal classification.

The impact of class imbalance correction on model discriminative performance and probability calibration in clinical risk prediction remains unclear. This study systematically evaluates the effects of SMOTE, random oversampling (ROS), and random undersampling (RUS) across ten real-world clinical datasets using a range of linear and nonlinear models. Comprehensive comparisons are conducted using metrics including ROC-AUC, Brier score, and calibration intercept/slope. Results indicate that none of the three resampling methods significantly improve discrimination, yet all consistently degrade probability calibration—evidenced by increased Brier scores (0.029–0.080) and substantial shifts in calibration parameters—revealing systematic distortion in predicted risk estimates. These findings challenge the conventional use of resampling techniques in clinical prediction modeling.

This work addresses the lack of general-purpose methods and open-source tools for ordinal classification by proposing a model-agnostic framework that transforms any base classifier into an ordinal-aware variant. The approach integrates a classifier pooling strategy with ordinal constraint mechanisms, enabling, for the first time, universal adaptation of arbitrary classifiers to ordinal data. To support reproducibility and practical adoption, the authors release an open-source Python package that fills a critical gap in available ordinal classification tooling. Extensive experiments on multiple real-world datasets demonstrate that the proposed method significantly outperforms conventional non-ordinal classifiers, particularly in small-sample and high-cardinality settings, thereby confirming its effectiveness and practical utility.

This work addresses the need for uncertainty quantification in ordinal classification within high-stakes domains such as medicine and finance, where errors of varying severity must be rigorously controlled. Existing conformal prediction methods are limited by their choice of nonconformity functions, which often fail to reflect the inherent ordering of classes. To overcome this, the authors propose a novel conformal prediction approach based on the Ranked Probability Score (RPS), introducing RPS as a natural nonconformity measure that captures ordinal risk. This method yields continuous prediction sets centered around the median, avoids greedy search procedures, and maintains model-agnosticism and computational efficiency. It is applicable to both evaluation-based and grouping-based ordinal tasks. Empirical results across multiple image and tabular ordinal datasets demonstrate that the proposed method achieves a superior trade-off between prediction set width and the severity of miscoverage compared to existing approaches.

Existing conformal prediction methods for ordinal classification often rely on heuristic designs or strong distributional assumptions (e.g., unimodality), compromising model- and distribution-freedom, and lack systematic analysis of the coverage-efficiency trade-off. This work proposes the first model-agnostic and distribution-agnostic framework for well-calibrated ordinal prediction, formulating ordinal classification as an instance-wise minimum-length coverage problem. We design a linear-time sliding-window algorithm to construct locally optimal prediction sets and introduce length regularization to enhance set compactness. Evaluated on four cross-domain benchmark datasets, our method reduces average prediction set size by 15% compared to state-of-the-art approaches, demonstrating superior efficiency, robustness, and generalization capability while maintaining rigorous marginal calibration.

This work addresses the challenge that class labels in medical prediction tasks often exhibit clinically meaningful ordinal structures, which standard loss functions fail to account for by treating all misclassifications equally. To this end, the authors propose the Ordinal Cross-Entropy (OCE) framework, which uniquely integrates an asymmetric, distance-sensitive ordinal cost matrix into the cross-entropy loss. This formulation preserves probabilistic interpretability while enhancing optimization stability and ordinal consistency. The method is fully differentiable and amenable to end-to-end training with deep neural networks. Extensive experiments on multiple medical benchmark datasets demonstrate that OCE significantly reduces misclassification costs and improves prediction calibration, outperforming current state-of-the-art ordinal regression approaches.

This study addresses the lack of systematic comparison between cross-validation and bootstrapping for assessing instability in clinical prediction models. Leveraging a cohort of 19,418 emergency department patients, it presents the first comprehensive evaluation of repeated five-fold cross-validation versus bootstrapping across varying events-per-variable (EPV) scenarios, using logistic regression and random forest models. Performance was assessed via AUC, calibration slope, large-scale calibration, and mean absolute prediction error (MAPE). Results indicate that when EPV ≥ 30, both methods yield comparable discriminative ability; however, cross-validation provides more accurate calibration estimates and significantly lower MAPE. These advantages render cross-validation particularly suitable for evaluating model instability across multiple algorithms, offering a dual benefit of internal validation and quantification of predictive stability.

This work addresses the lack of instance-level criteria for determining when noisy labels generated by vision-language models (VLMs) in weakly supervised learning transition from beneficial to detrimental. It introduces the first instance-level calibration method grounded in classical label noise theory, adapted for modern foundation models. By calibrating BiomedCLIP-generated weak labels using only 10–20 gold-standard annotations on three medical imaging benchmarks—PCAM, ISIC, and NIH-CXR—the study identifies performance crossover points (approximately 100, 20–50, and 250–500 samples, respectively), beyond which downstream AUC drops by up to 0.10. The findings reveal that the labeler, rather than the student model, dominates performance, and propose label space projection as a theoretically motivated improvement. The resulting decision rule demonstrates robustness across diverse pretrained architectures.

Whether class imbalance correction improves the performance of clinical prediction models remains controversial. This study leverages data from the GUSTO-I clinical trial to systematically evaluate the impact of various correction strategies—including algorithm-level rebalancing, oversampling, and hybrid sampling—on model discrimination (AUC), calibration (calibration plots and MAPE), and predictive stability (Classification Instability Index, CII) across varying sample sizes. Using penalized logistic regression with 200 bootstrap replications, we find that all correction methods fail to enhance discriminative performance and instead introduce greater calibration bias, risk overestimation, and increased prediction instability. These results challenge the common practice of routinely applying class imbalance corrections in clinical modeling and, for the first time in large-scale simulations, reveal their potential harms.

This study addresses the unification of calibration concepts across classification and regression tasks, aiming to ensure consistency between predicted distributions and observed outcomes for diverse data types—continuous, discrete, nominal, and binary. The work introduces modal calibration for nominal outcomes and establishes a hierarchical framework distinguishing full, partial, and average calibration. It proposes a generalized definition of calibration based on predictive distribution functionals—such as means, quantiles, and event probabilities—and leverages probability integral transforms alongside constructive algorithms for analysis. Key contributions include demonstrating the logical independence between dual probability integral transform (PIT) calibration and existing discrete calibration notions, clarifying implication and independence relationships among various calibration types, and providing reproducible methods for generating illustrative examples and counterexamples.